Research map / Adversarial robustness and security / Black-box and transfer attacks
Towards More Practical Adversarial Attacks on Graph Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use greedy correction of random-walk column-sum scores instead of Degree node ranking | black-box GNN feature attack · Cora, JKNetConcat, threshold 10% | accuracy (%): 60.7 -> 52.7 | Helped |
| Use greedy correction of random-walk column-sum scores instead of Degree node ranking | black-box GNN feature attack · Citeseer, GCN, threshold 10% | accuracy (%): 67.5 -> 65.1 | Helped |
| Use greedy correction of random-walk column-sum scores instead of Degree node ranking | black-box GNN feature attack · Pubmed, JKNetMaxpool, threshold 10% | accuracy (%): 63.3 -> 60.6 | Helped |
| Use greedy correction of random-walk column-sum scores rather than random node selection | black-box GNN feature attack · Cora, JKNetConcat, threshold 10% | accuracy decrease (%): 17.4 -> 33.5 | Helped |
| Use GC-RWCS rather than RWCS importance ranking | GCN attack on Citeseer · Citeseer, GCN, varying perturbation strength | classification accuracy | Helped |
| Use RWCS rather than centrality metrics | GCN attack on Citeseer · Citeseer, GCN, varying perturbation strength | classification loss | Helped |
| Increase perturbation strength for RWCS | GCN attack on Citeseer · Citeseer, GCN, varying perturbation strength | mis-classification rate | No clear effect |
| Vary GC-RWCS random-walk step count across L=3,4,5,6,7 (ablation) | GC-RWCS · Cora, Citeseer, and Pubmed; GCN and JK-Nets | accuracy (%): typically within 2% | No clear effect |
| Apply GC-RWCS node selection to GAT rather than only GCN-derived setups | GAT feature attack · Cora, Citeseer, and Pubmed, threshold 10% | accuracy (%) | Helped |
Closest papers by meaning
- Highly Imperceptible Black-Box Graph Injection Attacks with Reinforcement Learning (AAAI 2025 · similarity 0.90)
- Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.89)
- Adversarial Robust Generalization of Graph Neural Networks (ICML 2025 · similarity 0.89)
- Deterministic Certification of Graph Neural Networks against Graph Poisoning Attacks with Arbitrary Perturbations (CVPR 2025 · similarity 0.89)
- A Simple and Yet Fairly Effective Defense for Graph Neural Networks (AAAI 2024 · similarity 0.89)
- Bounding the Expected Robustness of Graph Neural Networks Subject to Node Feature Attacks (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.